2012-09-30 Martin Maechler
* R/dpqr-stable.R (dstable, .fct1): direct 'log' argument;
advantageous in few (only) cases. New default 'zeta.tol=NULL' now
determines zeta.tol depending on parameters.
2012-09-29 Martin Maechler
* R/dpqr-stable.R (dstable.smallA): new asymptotic (alpha -> 0)
formula for small alpha.
* DESCRIPTION (Version): 0.6-5, released to CRAN on 2012-10-01
2012-05-29 Martin Maechler
* DESCRIPTION (Version): 0.6-4, released to CRAN on 2012-05-29
2012-03-19 Martin Maechler
* DESCRIPTION (Version): 0.6-3, released to CRAN on 2012-03-19
2011-03-29 Martin Maechler
* R/dpqr-stable.R (dstable): new 'zeta.tol' argument; and checking
for "absolut+relative" closeness to 'zeta'.
* R/dpqr-stable.R (dstable, pstable): lower the default 'tol' to
from 16 * epsC to 64* epsC.
2011-03-29 Martin Maechler
* DESCRIPTION (Version): 0.6-2.1, released to CRAN on 2011-03-30
* R/dpqr-stable.R (dstable(*, alpha=1) -> .fct2): "robustified" g()
* R/dpqr-stable.R (pstable(*, alpha=1) -> .FCT2): ditto + simplified (1 integ.)
2011-03-28 Martin Maechler
* DESCRIPTION (Version): 0.6-2
* R/dpqr-stable.R (dstable -> .fct1): another round of
improvements, thanks to Dieter Schell's bug reports.
2011-03-24 Martin Maechler
* R/dpqr-stable.R (qstable): also add arguments lower.tail, log.p
2011-03-23 Martin Maechler
* DESCRIPTION (Version): 0.6-1.1
* R/dpqr-stable.R (dstable -> .fct1 -> g): more careful computation
of g(), notably at +- pi/2 where it could become NaN instead of 0.
Numerically better but considerably (~ 30-40% ?) slower!
* tests/dstab-ex.R: testing that
2011-03-21 Martin Maechler
* R/dpqr-stable.R (pstable): new arguments (lower.tail=TRUE, log.p=FALSE)
not used "properly" yet
(dPareto, pPareto): new auxiliary functions for tail behavior; not
used yet - apart from tests:
* tests/tails.R: new: testing tail behavior.
* R/dpqr-stable.R (qstable): infinite while(.) now caught
2011-03-12 Martin Maechler
* DESCRIPTION (Version): 0.6-0, released to CRAN, 2011-03-14
* R/dpqr-stable.R (dstable, .fct1): detect when the integration
seems completely unreliable and then also use the x=zeta value.
This cures all remaining big problems I've found.
2011-03-10 Martin Maechler
* R/dpqr-stable.R (dstable, .fct1): use uniroot() to find theta2; then
possibly split into four (instead just two) intervals for integration.
2011-03-08 Martin Maechler
* R/dist-stableMode.R (stableMode): add 'tol' (and 'beta.max')
arguments which were implicit before.
* R/utils.R (.unirootNA): add 'tol' and 'maxiter' arguments.
* R/dpqr-stable.R (pstable): get rid of "Sversion" and add 'tol,
subdivisions' as arguments.
Slightly *increase* default tol(erance) - so .integrate2() no
longer warns, at least not for the examples.
2011-02-02 Martin Maechler
* DESCRIPTION (Package): renamed to 'stabledist' as Jim Lindsey has
maintained a non-CRAN package 'stable' forever (and it's used).
* TODO: added